Classification of Learning Profile Based on Categories of Student Preferences

Classification of Learning Profile Based on Categories of Student s Luciana A M Zaina, Graça Bressan University of São Paulo, lzaina@larc.usp.br,gbressan@larc.usp.br Abstract - In an environment applied in engineering teaching, as in many knowledge areas, is very important to know and understand learner differences in a way to be able to adapt system s actions to student s best learning conditions and aptitudes. Working thus makes it possible to identify learning profiles within a group of students, allowing the system to supply learners with contents and tools more suited for them. The goal of this work is to present the architecture of a system that realizes an evaluation of learning profiles based on categories of student preferences. The categories are defined from Felder-Silverman Learning Style Model. The architecture enables the teacher to specify the observable characteristics he considers most suitable within the teaching scope in question, whose characteristics are related with categories of student preferences. Through the categories create a relationship between what is observed and the learning objects used to build automatically the learning scenarios according to the student learning profile. Index Terms learning profile, learning styles, observable characteristics. INTRODUCTION The need to adapt teaching strategies to the student s preferences is a reality in classrooms, be they physical or virtual. However, this does not mean that a method should be created for each student in a classroom, but that the best form of interaction for each of them be identified, building groups of learners with common characteristics [2]. The personalization of a learning process occurs through the investigation of the student s characteristics. Based on information obtained from his explicit and implicit knowledge, it is possible to model his needs within the context in which he interacts [3]-[4]. The student model supports the adaptation learning process in a specific environment; because of this the decision about the student model content is primarily related with the learning process goal. A student model might be an approximate, possibly partial, representation of student preferences about a particular domain that fully or partially accounts for specific aspects of student behavior. The common explored data about the student is his characteristic, preferences and skills. Besides of this, it is very import to know the context of student interaction to make the relationship between the learning and the scenario where it occurs [8]-[10]. Observing the student interaction is one the most important phase of adaptation process. Because of this, it is necessary to adopt techniques and mechanisms to provide learners with different teach strategies to attend him in an individual way. One of these mechanisms may be to observe the learning style of the students [8]. The objective of this work is to present the architecture of a system which makes the classification of a student within a given learning profile based on profile models predefined by the teacher. The learning profile is identified from information about the student s interaction in a given e- learning environment, which is obtained from a given monitoring mechanism. The profile models are built according to categories of student preferences based on the proposal of learning styles put forward by Felder and Silverman [4], [5]. A relationship between the categories of preferences and the learning objects is used to build automatically the learning scenarios according to the student learning profile. This architecture was defined with the main purpose of being applied to the teaching of engineering. LEARNING PROFILE AND LEARNING STYLES The learning style involves the strategies that a student tends to apply frequently to a given teaching situation. Each individual can fit into different styles that cause him to adopt attitudes and behavior that are repeated in different moments and situations. The learning style of a student evidences his/her learning profile. The learning profile defines a set of characteristics that classifies a student in a specific model during his learning process [6], [14]-[15]. Learning styles are cognitive, affective and psychological traits that determine how a student interacts and reacts in a learning environment [6]. The idea is to identify the marked characteristics of a given learner so that these traits influence his learning process. However, to this end, it is necessary to pinpoint what one considers relevant to observe in the student s interaction with the learning environment [4]. To satisfy a given learning style, the teacher must use teaching strategies that can meet the needs of diverse learning perspectives. The idea is to identify the marked characteristics of a given learner so that these traits influence his learning process. However, to this end, it is necessary to F4E-1

pinpoint what one considers relevant to observe in the student s interaction with the learning environment [7]. There are numerous models for the implementation of learning styles, each of which is suitable for a different learning scope. According to Felder and Brent [6], although there are many models to be adopted as learning styles, five can be highlighted that are applicable in the area of engineering teaching: the Myers-Briggs Type Indicator MBTI, Kolb s Experiential Learning Model, the Hermann Brain Dominance Instrument (HBDI), the Dunn and Dunn Model, and the Felder-Silverman Model. Among these models, this work will adopt the one proposed by Felder and Silverman [5] to support the definition of profile models. The Felder-Silverman Learning Style Model is describe by dimensions of Learning and Teaching Styles, creating a relationship to learning styles and teaching strategies that could be adopted to support the student learning style [6]. Table I presents the dimensions of Felder-Silverman Learning Style Model (learning styles and their corresponding teaching strategies). The underlined words represent categories where the dimensions may be fit in. For example, the category delineates the best way to student understand contents based on his/her learning style. TABLE I DIMENSIONS OF FELDER-SILVERMAN LEARNING STYLE MODEL Learning Style sensory intuitive Teaching Strategies concrete abstract Features It is related with the perception of content. visual visual It is related with the format of auditory verbal content presentation. active reflective sequential global active passive sequential global It is related with the student participation in the activities. It is related with the best order to present the content: step-by-step progression or a overview first of content. According to categories and the dimensions of the model presented in Table I it will be possible to create models of learning profile. The Felder-Silverman model was selected, because it s close relationship to learning styles and teaching strategies, resulting in an adherence between these aspects. In [5], Felder and Silverman pinpoint the application of this model in Engineering Education. This is an important point, because there is the intention to apply the architecture proposed in this area. ARCHITECTURE PROPOSED The need to adapt educational environments to teaching strategies that fit a student s learning profile has become a widely discussed issue. Based on this need, this paper proposes architecture of a system that observes the student in a learning process, identifying the learning profile to which a given student is best fitted. The learning profile is obtained through the analysis of data from the student s interaction, based on profile models previously define by the teacher. Figure 1 shows an overview of the architecture which is composed to modules. The architecture enables the teacher to specify the observation features he considers most suitable within the teaching scope in question. The possibility of modeling observation features affords the environment greater flexibility for evaluating learning profiles. This can be done in Module of Observation Planning. FIGURE 1 ARCHITECTURE PROPOSED MODELING OBSERVABLE FEATURES Before starting the definition of observable features is necessary to specify what model of learning style should be used. Various combinations can be derived from the topics defined in [5] to identify and classify the student according to a given learning style. The architecture allows the teacher to define learning profile models which he/she wishes to observe in student behavior and to suggest contents based on profile evaluation. A question to be answered is which data of the student s interaction should be taken into account. In fact, the data are dependent on the teaching objective. For this reason, the classification of the student s learning profile cannot be the same in different teaching scopes. It is necessary to define the features that will be observed during the student interaction in the electronic environment. The teacher will determine the relevant observable features in the Module of Observation Planning. This work adopts two classifications for the observable information about the learning profile: Check Points (insertions of tests and exercises that enable the teacher to make point checks to track the knowledge construction by the student, or even about his familiarity with and/or preference for the adoption of a given type of material or tool); and Learner Interactions (which represent the learner s interactions with the environment, his preferences for types of materials, number of accesses to a given content, interactions with a tool, ). The observation planning should be composed by the follow information: Description of observable feature: describe what will be observed. Function: each observable feature can have a function associated to it, whose purpose is to consolidate the information related to the observable feature. The F4E-2 Student Electronic Environment Locator of learning objects Monitoring Module Learning profile suggested Module of Observation Planning Profile Evaluator Module

functions used are functions that already exist in the (Learning Profile Evaluation System) functions library. Location of observation: where the observation will be occurred in the learning environment. Moment of observation: when the observation must be done. Table II presents two observable features of learner interaction type. feature Media Material TABLE II EXAMPLE OF OBSERVABLE FEATURES Function Location Moment MediaMore MaterialType Moreo environment environment interaction interaction The Media observable feature has the goal to verify the format of media that the student more accessed. One student may access different media during his/her interaction in the environment. The MediaMore function has the purpose of consolidate all of these events caught during the student interaction, confronting not only the type of media, but the time that the student spends using it. The consolidation resulting is the format of the media which the student more accessed. One observable feature will reflect a student preference about the feature. Because of this the teacher must classify, during the observation planning, each observable feature in one of categories of preferences:, Presentation Format, Presentation Order or Participation. These categories are adherent to dimensions of Felder-Silverman Learning Style Model presented in Table I. This means that an observable feature classified as will reflect a sensory or intuitive learning profile. Considering the example of Table II is possible to classify as follow (Table III): TABLE III EXAMPLE OF OBSERVABLE FEATURES CLASSIFICATION feature Presentation Format Media Material The group of observable features will compose an Observation Model. The Observation Model will send to Monitoring Module to be used by tracking the student interaction in the e-learning environment. MODELING LEARNING PROFILES The next step is the values specification for each observed feature determining the learning profile types that will be adopted during classification process. The values permit the system to distinguish the different types of learning profile considering the variety of observable features. For example, when the observable item Media has the video value the system consider that the learner should be fitted in a determined learning profile. The teacher should specify values to each observable feature for defining each type of learning profile he will adopt. Table IV and Table V defines two type of learning profiles: TABLE IV EXAMPLE OF TYPE OF LEARNING PROFILE : VISUAL-CONCRETE Type of Learning Profile: Visual-Concrete Presentation Format: Visual Characteristics of Learning Profile: feature : Concrete feature value Presentation Format Media Figure or Video Material Examples or simulations or problem resolution TABLE V EXAMPLE OF TYPE OF LEARNING PROFILE : VERBAL-REFLEXIVE Type of Learning Profile: Verbal-Abstract Presentation Format: Verbal Characteristics of Learning Profile: feature : Abstract feature value Presentation Format Media Text or Sound Material Expositive texts or articles The first example (Table IV) defines observable feature values that describe a Visual-Concrete type of profile. When the student fits in observable features of this type he/she may be classified as Visual-Concrete. The teacher may specify the characteristics of each type of learning profile for the categories of preference used in the observed feature definition. Table IV describes Visual value to the category Presentation format and Concrete value to the category, for instance. One type of learning profile may have many observable features which will be evaluated together during student profile classification process. PUBLISHING OF OBSERVATION MODEL After the definition of observable features and the learning profile types, creates the repository to store information from the monitoring module. Besides this, it sends the Observation Model to monitoring module to track the student interaction catching information planned (Figure 2). MONITORING MODULE The monitoring module catches the results from student interaction of observable features and it sends to that stores the results in the repository. There is no analysis regarding learner interaction during the monitoring process, F4E-3

because it is important to get a variety of data to analyze different actions of the student [8]-[9]. The analysis of data from the tracking of student interaction will occur when a student completes a teaching module. Electronic Environment Monitoring Module Repository sends FIGURE 2 ARCHITECTURE PROPOSED PROFILE EVALUATOR MODULE Observed features creates When a student completes a teaching module, the monitoring module triggers an event to, notifying it of the conclusion of the process and informing it who is involved in the interaction. Based on this information, the Profile Evaluator Module consolidates the information about the student s interactions (using the functions associated to each observable feature) that is contained in the repository. The result of this consolidation will determine the values of each item described in the observable feature for a specific student, thus providing information to determine the student profile. Upon concluding the consolidation, the Profile Evaluator Module executes a profile classification algorithm. then classifies the given student according to the learning profile (defined by the teacher previously) he best fits into. Because the profile is evaluated based on several observable features, it is very likely that a learning profile cannot be found that has exactly the same values as those observed for a given learner. To solve this problem, the profile evaluator adopts an instance-based learning algorithm called 1-Nearest Neighbor. Briefly, the 1-Nearest Neighbor algorithm finds the objects that make up a set of systemically known examples closest to another object that is being observed at a given moment [1], [13]. In this work, the set of examples is given by the types of learning profiles defined during the modeling. First is checked the distance between one observable feature (observed value obtained during the monitoring process) and its correspondent feature in a type of learning profile. This distance is called featuresdistance. If the features have correspondence the featuresdistance has the value 0. Otherwise the featuresdistance has the value 1. Figure 3 shows an example of featuresdistance analyses. According to monitoring results and the types of learning profile, it is possible to note that featuresdistance is 0 between the observed value and the observable feature value of Visual-Concrete learning profile. The total distance will be calculated after the featuresdistance calculation for each observable feature. The distance between the analyzed profile and a specific type of learning profile, called total distance is obtained from the sum of all the featuresdistance. For example, the total distance between the analyzed profile and the Visual- Concrete profile is acquired from the sum of all the featuresdistance of the respective type of profile. The equation (1) represents the calculation of total distance : totaldis tan ce( PT, AP) = n feturesdis i PT = profile type AP = analyzed profile i = observable feature n = number of observable features analyzed tan ce After the calculation of the totaldistance between the analyzed profile and each profile type, the algorithm chooses the type of profile that has the minor totaldistance. Thus determines the suggested profile of the student. Presentation Format Set of Types of Learning Profile feature Media Observed feature Value Video Media Monitoring results feature Visual-Concrete featuresdistance FIGURE 3 EXAMPLE OF FEARTURESDISTANCE CALCULATION Figure 4 illustrates the dynamic of to recommend the learning profile. CATOR OF LEARNING OBJECTS Visual- Concrete feature value Figure or Video Verbal- Abstract feature value Text or Sound Verbal-Abstract featuresdistance Media 0 1 Calculate of featuresdistance (1) A learning object can be defined as an entity that can be applied in a teaching-learning process. It may be a text, a video, a figure, a simulator, Within the scope of electronic learning, the aim is to create contents in digital form that can be reused for different learning objectives or even employed in the construction of other learning objects [12]. F4E-4

Monitoring Module Who Profile Evaluator Module Consolidation of information of the fields: Title, Description, and Keywords. The result of the search will be the first selection of the possible objects that may be presented to the student. Steps to localize learning objects M fields M Category Repository Classification of Learner Profile Location of concepts Title, Description, and Keywords General Suggested Learning Profile FIGURE 4 DYNAMIC OF LEARNPES One of the ways to organize learning objects so that they can be reused and utilized systematically is through the description of metadata. A metadata can be defined as a set of attributes that describes an entity, facilitating the understanding, use and management of this entity. One of the specifications of the most commonly used metadata is done through the M (Learning Object Metadata) standard of the Institute of Electrical and Electronics Engineers IEEE. The M standard [11] has a structure that describes learning objects through descriptor categories that detail data on a given learning object. Each category has a specific purpose, such as describing general attributes of objects, educational objectives, Profile type planning involves the specification of the categories of preferences related with these types of profiles. For example, Table IV specifies that the categories used are Presentation Format and. The fact that the type of profile was defined as Visual and Concrete means that it is characterized by these definitions within the categories of Presentation Format and, respectively. These categories are utilized to locate the learning objects that will match the profile suggested by the environment, relating the categories of preferences to the fields that specify the learning objects according to the M (Learning Object Metadata) standard. This work proposes that the learning objects can be localized so that they adhere to the learning profile and the concepts one wishes to broach in a specific learning scenario. The location is done through the Learning Object Locator, which carries out searches in repositories containing objects catalogued according to the M standard. Figure 5 presents the learning objects selection process. The maintenance of learning objects repositories must be supported by the learning infrastructure that adopts the proposed architecture. When a leaning object is catalogued it will be store in the repository that adopts M specification. The Learning Object Locator will use the description of the concepts of the topics relating to the learning content in order to conduct the first location step. To this end, it uses the General category of the M specification by means Finding the objects that match the student s learning profile Learning Objects selected Interactivity and Learning Resource FIGURE 5 LEARNING OBJECTS SELECTION PROCESS The second location step consists of finding the objects that match the student s learning profile in the set of objects selected in the first step. This is done by verifying the student s learning profile. Using the Educational category of the M standard (Interactivity and Learning Resource fields), the Locator will locate the objects that match the preferences related to the student s learning profile from the set of objects obtained in the first locating step. The Locator will use the Table VI in order to find the objects related to the student s learning profile. When the characteristics of a student s profile indicate he is active or reflective, the teaching strategies that can be adopted are forums, chats, group discussions, For example, if a student has a profile Visual- Concrete as his/her learning profile, the Locator of Learning Objects will search in Interactivity and Learning Resource fields the objects to attend this profile. The locator will associated the characteristic Visual with figures, videos, movies, After carrying out the second location step, the electronic learning environment will receive all the selected learning objects to present them to the student. LEARNING SCENARIO Educational A learning scenario may be considered an entity composed by a set of elements which supports the student during the learning process. The scenario must concern information about the teaching core as the modules of a course, tools, links of material and so one. Concisely, a learning scenario includes the essential items to achieve the learning goal. The items are necessary units to construct the learning process in a gradual cadence. The learning scenarios will be built based on learning objects selected by the Locator of Learning Objects. The electronic learning environment may present more than one learning object to the student allowing him to select the learning object he wishes to work. F4E-5

TABLE VI RELATIONSHIP BETWEEN MM FIELDS AND PROFILE CHARACTERISTICS M field Type of Interactivity Type of Learning Resource Description of the field s function Describes objects with concrete characteristics (simulation, exercises, case studies, ). Describes objects with abstract characteristics (expository contents) Indicates if the object is a figure, diagram, sound, Indicates if the object is a practical exercise, a case study, Indicates if the object is a complementary reading, theoretical questionnaire, Value of the field active expositor y Figure Video Movie, Text, Sound, Practical exercise, Experime nt, Question naire, Reading CONCLUSIONS Characteristi c of the Profile Concrete Abstract Visual Auditory Active Reflective the Profile Presentation Format Student Participation The development of flexible educational environments (which can be used in different areas of education) that are adaptable has become an important requisite within the teaching-learning process. The creation of different learning scenarios considering different learning profiles is an essential aspect to meet the individual needs of students in an e-learning environment. Within this scope, the identification of learning profiles is indispensable to allow for adequate adaptations to be done. The architecture proposes a model for the definition of learning profiles so that these profiles consider the student s interaction during the period of observation. In addition, the proposed architecture classifies the student within a given learning profile, analyzing the observable features jointly, thereby allowing for profiles more suitable to a given situation to be suggested. The association between learning profiles and learning objects metadata grants dynamism in the process of learning scenarios construction. One future work is to apply the proposed architecture in a group of Computer Engineering students to realize the analyses over the selected learners profiles. From this experience will be possible to verify the convergence of observed features analyzed and the real student profile. After that, the architecture will be expanded to adopt feature distance not only 0 or 1, considering the other possible values that a feature can accept during the interaction. ACKNOWLEDGMENT Our thanks to FAPESP for sponsoring this work. REFERENCES [1] Aha, W, D D. Kibler, and M. K. Albert, Instance-Based Learning Algorithms, Machine Learning, 6, 1, 1991, pp 37-66. [2] Akhras, N, F, and Self, A, J, Modeling the Process, not the Product of Learning, Computer as Cognitive Tools No More Walls, pp 3-28. [3] Brusilovsky, P., and Peylo, C, Adaptive and intelligent Web-based educational systems, International Journal of Artificial Intelligence in Education, 13, 2-4, 2003, pp. 159-172. [4] Coffield, F. et al, Learning styles and pedagogy in post-16 learning. A systematic and critical review, Learning and Skills Research Centre, London, 2004. [5] Felder, M, R, and Silverman, K, L, Learning and Teaching Styles in Engineering Education, Journal of Engineering Education, 78, 7, 1988, pp. 674-681. [6] Felder, M, R, and Brent, R, Understanding Student Differences, Journal of Engineering Education, 94, 1, 2005, pp 57-72. [7] Fenrich, P, Practical Suggestions for Addressing Learning Styles in Computer-Based Simulations, Proceedings of The Ninth IASTED International Conference on Computers and Advanced Technology in Education, Lima, Peru, October, 2006, pp 61-66. [8] Fisher, G, User Modeling in Human-Computer Interaction, User Modeling and User-Adapted Interaction, 11 (1/2), 2001, pp 65-86. [9] Kobsa, A, Generic User Modeling Systems, User Modeling and User-Adapted Interaction, 11, 1/2, 2001, pp. 49-63. [10] Kobsa, A, Koenemann, J, and Pohl, W, Personalised hypermedia presentation techniques for improving online customer relationships., The Knowledge Engineering Review, 16 (2), 2001, pp 111 155. [11] M, Draft standard for learning object metadata, 2002. http://ltsc.ieee.org.wg12/index.html. [12] McGreal, R, Learning Objects: A Practical definition, International Journal of Instructional Technology and Distance Learning, 1, 9, 2004. [13] Silva, T, G, and Rosatelli, C, M, Adaptation in Educational Hypermedia based on the Classification of the User Profile, Proceedings of the 9th International Conference on Intelligent Tutoring Systems (ITS 2006), 2006, pp 268-277. [14] Stash, N, Cristea, A, and De Bra, P, Authoring of Learning Styles in Adaptive Hypermedia: Problems and Solutions, Proceedings of the WWW 2004 Conference, 2004, pp 114-123. [15] Tarpin-Bernard, F, Habieb-Mammar, H, Modeling Elementary Cognitive Abilities for Adaptive Hypermedia Presentation, User Modeling and User-Adapted Interaction, 15 (5), 2004, pp 459-495. AUTHOR INFORMATION Luciana A M Zaina Researcher, University of São Paulo, Brazil, lzaina@larc.usp.br. Graça Bressan, Professor, University of São Paulo, Brazil, gbressan@larc.usp.br. F4E-6